Pattern 1 — ECS Task Invoked by EventBridge (Event-Driven)

Trigger: Something happens in AWS (e.g. a file is uploaded to S3).

Client uploads file to S3
    |
S3 sends event to EventBridge
    |
EventBridge rule: "on S3 upload --> run ECS Task"
    |
ECS spins up a new Fargate task
    |
Task reads file from S3 --> processes it --> saves result to DynamoDB
    |
Task finishes and stops automatically

Key points:

• Task is not always running — it spins up only when the event fires
• ECS Task Role needs S3 (read) and DynamoDB (write) permissions
• Great for file processing, image resizing, data transformation

Pattern 2 — ECS Task Invoked by EventBridge Schedule (Cron Jobs)

Trigger: A fixed time interval — no human action needed.

EventBridge Schedule: every 1 hour
    |
Rule: Run ECS Task
    |
ECS spins up a new Fargate task
    |
Task does batch processing --> saves results to S3
    |
Task finishes and stops automatically

Key points:

• Runs on a schedule — nightly reports, hourly sync, daily cleanup, database backups
• Much cheaper than running a server 24/7 just to run a job once an hour

Pattern 3 — ECS + SQS Queue (Scale on Queue Depth)

Trigger: Number of messages waiting in an SQS queue.

Messages arrive in SQS Queue
    |
ECS tasks poll the queue for messages
    |
Queue backlog builds up
    |
CloudWatch monitors SQS queue depth metric
    |
ECS Auto Scaling adds more tasks
    |
More tasks = more consumers = queue drains faster
    |
Queue empties --> Auto Scaling removes extra tasks

Key points:

• Scaling metric is SQS queue depth — not CPU or memory
• Tasks poll the queue themselves — SQS does not push to tasks
• Perfect for order processing, image processing, notification sending